Colorectal cancer drug chemotherapy reaction prediction system and storage medium

Abstract

The invention relates to the field of multi-modal learning, in particular to a colorectal cancer drug chemotherapy reaction prediction system and a storage medium, and a computer program in the storage medium executes the following steps: constructing a PDO model and a PDOX model based on tumor cells; obtaining a standardized median inhibitory concentration value and a standardized relative tumor proliferation rate of the colorectal cancer patient by utilizing the PDO model and the PDOX model; based on the obtained medical record and pathological examination of the colorectal cancer patient, obtaining the age, ASA score, Ki-67 index, combined positive score and clinical outcome of the patient, and further constructing a sample set in combination with a standardized median inhibitory concentration value and a standardized relative tumor proliferation rate; and constructing a model for predicting the chemotherapy reaction of the colorectal cancer drug on the basis of multi-factor logistic regression, training and evaluating the model by using the sample set, and performing chemotherapy guidance on a to-be-treated colorectal cancer patient. The problems that in the prior art, chemotherapy reaction prediction of the colorectal cancer patient is not accurate enough and low in efficiency are solved.

Description

Technical Field

[0001] The present invention relates to the field of multimodal learning, and in particular to a colorectal cancer drug chemotherapy response prediction system and a storage medium. Background Art

[0002] Colorectal cancer (CRC), the third most common cancer worldwide, poses a serious threat to human health. Despite the diverse treatment options currently available for CRC, the complex mechanisms of tumorigenesis, heterogeneity, and recurrence and metastasis result in significant variability in clinical responses to chemotherapy, radiotherapy, and their combination. For patients diagnosed with stage IV CRC, 5-fluorouracil, 5-fluorouracil + oxaliplatin, and 5-fluorouracil + irinotecan are commonly used chemotherapy regimens. While these drugs can kill tumor cells and shrink tumors to a certain extent, they carry a high risk of drug resistance and severe side effects, making them often difficult for patients with advanced disease to tolerate. Therefore, developing accurate predictive models to screen the most appropriate drug regimens is crucial for reducing the treatment burden on patients, preventing tumor progression, prolonging survival, and improving quality of life.

[0003] In clinical practice, treatment decisions for patients with stage IV colorectal cancer face numerous challenges. While the choice of chemotherapy regimen must be considered, the patient's clinically acceptable time window is crucial. This window, constrained by the patient's physical condition, reflects the timeframe within which they can receive treatment. Given the severity and poor physical condition of stage IV patients, chemotherapy regimens must be designed to balance tolerability and treatment urgency. Rationally scheduling chemotherapy cycles and doses within the clinically acceptable time window can both reduce the patient's burden and enhance treatment efficacy. Therefore, optimizing chemotherapy regimens by comprehensively considering individual patient differences, tumor biology, and treatment tolerance is key to prolonging patient survival and improving quality of life.

[0004] The emergence of patient-derived tumor organoid (PDO) and patient-derived tumor organoid xenograft (PDOX) models has opened new avenues for colorectal cancer treatment. PDO models are three-dimensional cell clusters formed by culturing tumor tissue in vitro. They mimic in vivo tumor growth and function and can be derived directly from patient tumor tissue. Compared to traditional two-dimensional cell culture, PDO models more closely resemble the human physiological environment, enabling more accurate assessment of drug efficacy and toxicity, improving the success rate of drug development, and can also be used for high-throughput drug screening to accelerate new drug development. PDOX models are humanized xenograft models established by transplanting PDO into immunodeficient mice. These models not only retain a high degree of similarity to the primary tumor in terms of histopathology, biomarkers, genetic characteristics, and pharmacological function, but also facilitate in vivo pharmacological testing.

[0005] However, existing prediction schemes based on these models have obvious flaws. The establishment of a single PDX model takes a long time, from 6 months to 2 years. For stage IV patients in urgent need of treatment, this may lead to missing the best treatment opportunity. In addition, its success rate varies greatly between different tumor types and patients, and some patients cannot obtain effective predictive information from it. The single PDO model lacks the components of the in vivo microenvironment and cannot evaluate in vivo pharmacokinetics. It is difficult to fully simulate the real situation in the body when predicting drug response. Although the PDO-PDX combined model combines the advantages of both, the establishment process is complex, time-consuming, and costly, and data integration is difficult, which seriously limits its widespread application in clinical practice. Therefore, it is urgent to develop a more efficient, accurate and practical prediction model. Summary of the Invention

[0006] In response to the defects in the existing technology, the present invention provides a colorectal cancer drug chemotherapy response prediction system and storage medium, which solves the problem of inaccurate and low efficiency in predicting chemotherapy response of colorectal cancer patients in the existing technology.

[0007] To achieve the above-mentioned object, one aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by a processor, the processor performs the following steps: obtaining tumor cells of a colorectal cancer patient, and constructing a PDO model and a PDOX model based on the tumor cells; based on drug chemotherapy, obtaining a standardized half-maximal inhibitory concentration value of the colorectal cancer patient using the PDO model, and obtaining a standardized relative tumor growth rate of the colorectal cancer patient using the PDOX model; obtaining the medical records and pathological examination of the colorectal cancer patient, and constructing a PDO model and a PDOX model based on the medical records and ... the pathological examination of the colorectal cancer patient, and constructing a PDO model and a PDOX model based on the medical records and the PDOX model; The pathological examination obtains the patient's age, ASA score, Ki-67 index, combined positive score and clinical outcome; a sample set is established using the half-maximal inhibitory concentration value, the standardized relative tumor proliferation rate, the patient's age, the ASA score, the Ki-67 index, the combined positive score and the clinical outcome; a model for predicting colorectal cancer drug chemotherapy response is constructed based on multivariate logistic regression, the model for predicting colorectal cancer drug chemotherapy response is trained using the sample set, and the model for predicting colorectal cancer drug chemotherapy response after training is evaluated; based on the evaluation results, the colorectal cancer drug chemotherapy response model is used to provide chemotherapy guidance for patients with colorectal cancer to be treated.

[0008] The present invention constructs PDO and PDOX models by obtaining tumor cells from colorectal cancer patients, and then obtains standardized half-maximal inhibitory concentration values and standardized relative tumor proliferation rates. Combined with the patient's age, ASA score, Ki-67 index and combined positive score obtained from medical records and pathological examinations, a model for predicting colorectal cancer drug chemotherapy response is trained. This model accurately predicts chemotherapy response from multiple dimensions, comprehensively considers patient differences, overcomes the limitations of a single model, improves the stability and reliability of the prediction results, and thereby improves the accuracy of chemotherapy regimens for colorectal cancer patients.

[0009] Optionally, constructing a PDO model and a PDOX model based on the tumor cells includes: culturing the tumor cells in three dimensions in vitro to construct a PDO model that retains the pathological, genetic characteristics and drug response characteristics of the tumor cells; and transplanting the PDO model into immunodeficient mice to construct a PDOX model that simulates the in vivo tumor microenvironment.

[0010] The PDO model constructed through three-dimensional in vitro culture retains key tumor cell characteristics and can accurately simulate tumor drug responses in vitro, providing a reliable basis for drug sensitivity studies. The PDOX model constructed by transplanting the PDO model can simulate the in vivo tumor microenvironment, compensating for the lack of an in vivo microenvironment in the PDO model and enabling more realistic assessment of drug efficacy in vivo.

[0011] Optionally, the method of obtaining the standardized half-inhibitory concentration value of the colorectal cancer patient using the PDO model includes: conducting experiments on the drug chemotherapy with different concentrations on the PDO model to obtain the actual measured half-inhibitory concentration value of the colorectal cancer patient; collecting historical half-inhibitory concentration values of colorectal cancer patients with the same drug chemotherapy; obtaining the median half-inhibitory concentration value of the drug chemotherapy based on the historical half-inhibitory concentration values; and calculating the standardized half-inhibitory concentration value of the colorectal cancer patient based on the actual measured half-inhibitory concentration value and the median half-inhibitory concentration value.

[0012] By conducting chemotherapy experiments with different drug concentrations on the PDO model, the present invention directly obtains actual patient measurements, reflecting individual drug sensitivity. Collecting historical data and calculating the median can serve as a standardization benchmark. Calculating the standardized inhibitory concentration (SI50) based on these two values effectively eliminates the effects of varying experimental conditions and sample variability, making data comparable across patients and improving the usability of the standardized SI50 value.

[0013] Optionally, the standardized half-maximal inhibitory concentration value satisfies the following formula: in, is the normalized half-maximal inhibitory concentration value, To measure the actual half-maximal inhibitory concentration value, is the median half inhibitory concentration value.

[0014] The present invention effectively corrects data deviations by correlating actual measured values with medians, so that drug sensitivity data from different patients and experiments have a unified and comparable quantitative standard.

[0015] Optionally, the use of the PDOX model to obtain the standardized relative tumor proliferation rate of the colorectal cancer patient includes: dividing the immunodeficient mice corresponding to the PDOX model into an experimental group and a control group; measuring the tumor volume of the immunodeficient mice in the experimental group and the control group to obtain the initial tumor volume of the experimental group and the initial tumor volume of the control group; applying the drug chemotherapy to the experimental group and applying a solvent or placebo equal to the volume of the drug chemotherapy to the control group to obtain the post-experimental tumor volume of the experimental group and the post-experimental tumor volume of the control group; calculating the standardized relative tumor proliferation rate of the colorectal cancer patient based on the initial tumor volume of the experimental group, the initial tumor volume of the control group, the post-experimental tumor volume of the experimental group and the post-experimental tumor volume of the control group.

[0016] By dividing the experimental group into the control group, the present invention can effectively compare the drug effects, measure and calculate the tumor volume at different stages, accurately reflect the drug's inhibition of tumor proliferation in vivo, and improve the accuracy of the standardized relative tumor proliferation rate.

[0017] Optionally, the evaluation of the model for predicting colorectal cancer drug chemotherapy response after training includes: extracting the clinical outcome of each sample in the test set, and using the model for predicting colorectal cancer drug chemotherapy response to obtain the predicted probability of each sample; setting multiple classification thresholds, comparing the predicted probability with the classification threshold, and calculating the sensitivity and false positive rate under the classification threshold based on the comparison results and the clinical outcome; using sensitivity as the vertical axis and false positive rate as the horizontal axis, using the sensitivity and false positive rate under the classification threshold as coordinate points, and drawing the coordinate points into a curve; calculating the area value under the curve and the 95% confidence interval of the area value; setting a baseline threshold, comparing the difference between the area value and the baseline threshold through a two-sided test, and calculating the significance level of the difference; using the area value, the 95% confidence interval and the significance level to comprehensively evaluate the model for predicting colorectal cancer drug chemotherapy response.

[0018] The present invention can accurately measure the predictive ability of the colorectal cancer drug chemotherapy response prediction model for the actual chemotherapy response by extracting the clinical outcome of the sample and calculating the predicted probability. Setting multiple classification thresholds to calculate the sensitivity and false positive rate can comprehensively evaluate the performance of the colorectal cancer drug chemotherapy response prediction model under different judgment criteria, avoiding the limitations of a single threshold. Plotting the curve and calculating the area under the curve and the 95% confidence interval intuitively reflect the discrimination ability and stability of the colorectal cancer drug chemotherapy response prediction model. By comparing the area value with the benchmark threshold through a two-sided test and calculating the significance level, the effectiveness of the colorectal cancer drug chemotherapy response prediction model is strictly verified from a statistical point of view. Combining these evaluation indicators, the model performance can be judged comprehensively, objectively and accurately, thereby improving the reliability and accuracy of the colorectal cancer drug chemotherapy response prediction model in predicting colorectal cancer drug chemotherapy responses.

[0019] Optionally, the model for predicting colorectal cancer drug chemotherapy response is determined to be an available model when the comprehensive evaluation satisfies the conditions that the area value is greater than 0.5, the lower limit of the 95% confidence interval is greater than 0.5, and the significance level is less than 0.05.

[0020] The present invention can effectively screen out reliable models for predicting colorectal cancer chemotherapy responses, ensuring that the model's predictive ability is superior to random chance, with good stability and statistically significant differences.

[0021] Optionally, the model for predicting colorectal cancer chemotherapy response satisfies the following formula: in, To predict the probability of chemotherapy resistance, is the normalized relative tumor growth rate, is the standardized half-maximal inhibitory concentration value, is the patient's age, For ASA rating, is the Ki-67 index, is the combined positive score.

[0022] This model formula for predicting colorectal cancer chemotherapy response comprehensively considers multi-dimensional key factors such as standardized relative tumor proliferation rate and standardized half-maximal inhibitory concentration value, and can accurately quantify the probability of drug resistance.

[0023] Optionally, the use of the colorectal cancer drug chemotherapy response model to provide chemotherapy guidance for colorectal cancer patients includes: obtaining individualized data of the rectal cancer patients to be treated, and inputting the individualized data into the predicted rectal cancer drug chemotherapy response model to obtain the probability of drug resistance to the drug chemotherapy; setting a resistance probability threshold, comparing the resistance probability with the resistance probability threshold, and determining, based on the result of the comparison, that the rectal cancer patients to be treated are chemotherapy-sensitive populations; based on the fact that the rectal cancer patients to be treated are chemotherapy-sensitive populations, determining the physical tolerance of the rectal cancer patients to be treated to the drug chemotherapy in combination with the individualized data; and determining the cycle and dosage of the drug chemotherapy based on the physical tolerance.

[0024] The present invention obtains the probability of drug resistance by inputting individualized data, can accurately judge the patient's response tendency to chemotherapy drugs, set thresholds to distinguish chemotherapy-sensitive groups, make treatment more targeted, and determine the body's tolerance in combination with individualized data. The chemotherapy cycle and dosage are determined accordingly, fully considering the individual differences of patients. It can not only ensure the treatment effect, but also avoid excessive treatment that brings unnecessary burden to patients, realize personalized and precise treatment, improve patients' tolerance and compliance to chemotherapy, and thus improve the overall treatment quality of colorectal cancer and the quality of life of patients.

[0025] Another aspect of the present invention provides a colorectal cancer drug chemotherapy response prediction system, comprising an input device, a processor, an output device and a memory, wherein the input device, processor, output device and memory are interconnected, the memory comprises the computer-readable storage medium described in the previous aspect of the present invention, the memory is used to store a computer program, the computer program comprises program instructions, and the processor is configured to call the program instructions.

[0026] The colorectal cancer drug chemotherapy response prediction system of the present invention has a compact structure, stable performance, high integration and simple composition. It can stably execute the steps of the program instructions in a computer-readable storage medium provided in the previous aspect of the present invention, further improving the overall applicability and practical application capabilities of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 A flowchart of program instructions in a computer-readable storage medium according to an embodiment of the present invention; Figure 2 is a curve diagram showing the relationship between sensitivity and false positive rate according to an embodiment of the present invention; Figure 3 Schematic diagram of the structure of a colorectal cancer drug chemotherapy response prediction system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0028] Specific embodiments of the present invention will be described in detail below. It should be noted that the embodiments described herein are for illustrative purposes only and are not intended to limit the present invention. In the following description, numerous specific details are set forth to provide a thorough understanding of the present invention. However, it will be apparent to one of ordinary skill in the art that these specific details are not necessarily required to practice the present invention. In other instances, well-known circuits, software, or methods are not specifically described to avoid obscuring the present invention.

[0029] Throughout this specification, references to "one embodiment," "an embodiment," "an example," or "an example" mean that a particular feature, structure, or characteristic described in connection with the embodiment or example is included in at least one embodiment of the present invention. Therefore, appearances of the phrases "in one embodiment," "in an embodiment," "an example," or "an example" in various places throughout this specification are not necessarily all referring to the same embodiment or example. Furthermore, the particular features, structures, or characteristics may be combined in any suitable combinations and / or subcombinations in one or more embodiments or examples. Furthermore, those of ordinary skill in the art will appreciate that the figures provided herein are for illustrative purposes only and are not necessarily drawn to scale.

[0030] See Figure 1 In one embodiment of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by a processor, the processor performs the following steps: Step S1, obtaining tumor cells from colorectal cancer patients, and constructing a PDO model and a PDOX model based on the tumor cells.

[0031] In this embodiment, obtaining tumor cells from colorectal cancer patients must comply with medical ethics standards and obtain informed consent from the patients.

[0032] Specifically, if the patient undergoes surgery to remove the tumor, the surgeon will use a sterile scalpel, scissors, etc. to cut a 0.5-1 cubic centimeter tumor tissue sample from the typical area of the tumor during the operation.

[0033] If surgical access is not possible, a fine needle aspiration biopsy is performed under imaging guidance, using a specialized needle to percutaneously puncture the tumor and extract an appropriate amount of tumor cell tissue. The sample is immediately placed in a sterile container containing DMEM culture medium supplemented with various nutrients and antibiotics and quickly transported to the laboratory.

[0034] Next, within the laboratory's clean bench, the sample is washed with sterile PBS buffer to remove impurities such as blood and tissue debris. Fine tissue scissors are then used to cut the tumor tissue into small pieces approximately 1-2 mm in diameter. Enzymatic digestion is then performed using an appropriate amount of trypsin-EDTA digestion buffer, followed by digestion in a 37°C incubator for 1-2 hours. Following digestion, the tumor tissue is gently pipetted to disperse into a single-cell suspension. The suspension is then filtered through a cell sieve to remove incomplete tissue clumps, ultimately yielding pure, viable tumor cells.

[0035] The construction of the PDO model and the PDOX model based on the tumor cells specifically includes the following steps: Step S101 , culturing the tumor cells in three dimensions in vitro to construct a PDO model that retains the pathological, genetic characteristics, and drug response properties of the tumor cells.

[0036] In this embodiment, on a sterile operating table, the obtained tumor cells are first quickly transferred into a centrifuge tube containing a specific culture medium. The culture medium is a special culture medium designed specifically for tumor organoid culture and rich in various growth factors, extracellular matrix components, and antibiotics.

[0037] Then, gently pipette the cell suspension to evenly disperse it, then take an appropriate amount of cell suspension and mix it with Matrigel in a 1:2 ratio to form a cell-Matrigel mixture. Carefully inoculate the mixture into an ultra-low adsorption 96-well plate or other suitable culture container. The inoculation volume per well should be precisely controlled between 50 μl and 200 μl according to experimental requirements. After inoculation, place the culture plate in a constant temperature incubator at 37°C and 5% carbon dioxide for 30 minutes to 60 minutes to promote Matrigel solidification and provide a stable three-dimensional support structure for the cells.

[0038] After the Matrigel has completely solidified, slowly add fresh culture medium preheated to 37°C. The volume of culture medium should cover the cell-Matrigel mixture while avoiding disrupting the solidified structure. Remember to change the culture medium every 2-3 days to maintain the nutrient supply required for cell growth and remove metabolic waste.

[0039] During the culture process, the cells were regularly observed under a microscope to record changes in cell morphology, proliferation rate, and other information. After 7-14 days of culture, the tumor cells gradually proliferated and formed cell clusters with a three-dimensional structure, thus initially constructing the PDO model.

[0040] Finally, the constructed PDO model was quality tested, its pathological characteristics were observed using histopathological staining (such as hematoxylin-eosin staining), its genetic characteristics were detected using gene sequencing technology, and its drug response characteristics were verified through drug treatment experiments to ensure that the constructed PDO model retained the relevant characteristics of the original tumor cells.

[0041] Step S102: transplanting the PDO model into immunodeficient mice to construct a PDOX model that simulates the in vivo tumor microenvironment.

[0042] In this example, 4- to 5-week-old nude mice were selected and acclimated in an SPF environment for 3-5 days. Before surgery, the PDO model was rinsed three times with sterile PBS, the culture medium was removed, and the model was mixed with Matrigel to maintain its three-dimensional structure.

[0043] Mice were anesthetized by intraperitoneal injection of Avertin, and the back skin was disinfected before subcutaneous injection.

[0044] Subcutaneous tumor growth was monitored weekly after transplantation, and samples were harvested when the tumor volume reached 100 mm³. Pathological features were verified by HE staining, human tumor markers were analyzed by immunohistochemistry, and genetic signatures were detected using gene sequencing to ensure that the PDOX model retained the characteristics of the original tumor. This method simplifies the operational process and efficiently constructs a PDOX model that combines the characteristics of human tumors with the in vivo mouse microenvironment. This provides a reliable in vivo validation system for predicting chemotherapy response, overcomes the limitations of a single PDO model, and improves predictive accuracy and clinical practicality.

[0045] Step S2, based on drug chemotherapy, using the PDO model to obtain the standardized half-maximal inhibitory concentration value of the colorectal cancer patient, and using the PDOX model to obtain the standardized relative tumor proliferation rate of the colorectal cancer patient.

[0046] The drug chemotherapy refers to 5-FU alone, 5-FU+OX or 5-FU+IRI.

[0047] Wherein, the method of obtaining the standardized half-maximal inhibitory concentration value of the colorectal cancer patient using the PDO model specifically includes the following steps: Step S211 , performing chemotherapy experiments with different concentrations of the drug on the PDO model to obtain the actual measured half-maximal inhibitory concentration value of the colorectal cancer patient.

[0048] In this example, to obtain the 50% inhibitory concentration (PIC) value for drug chemotherapy in PDO models, the PDO model was first digested into organoid spheres and seeded at 50-100 cells / well in a 96-well plate for a pre-culture of 24 hours. Six gradient concentrations of chemotherapy drugs (including a solvent control) were then added. After replacing the culture medium, the cells were cultured for 48-72 hours, and cell viability was assessed using the CCK-8 assay. Absorbance was measured using a microplate reader, and the percentage of cell viability at each concentration was calculated. Dose-response curves were fitted using four-parameter logistic regression to determine the drug concentration that inhibited cell viability by 50%, representing the PIC value actually measured in colorectal cancer patients. The experiment was performed in triplicate and included controls to ensure data reliability.

[0049] Step S212, collecting historical half-inhibitory concentration values of colorectal cancer patients who received the same chemotherapy drug.

[0050] In this example, we first searched for stage IV patients who had received the same chemotherapy regimen (5-FU alone, 5-FU+OX, 5-FU+IRI) over the past three years through the hospital's electronic medical record system and biological sample library. We screened cases with pathologically confirmed cases, successful construction of a PDO model, and measurement of the half-maximal inhibitory concentration. Simultaneously, we collected similar data from other institutions and published studies through a multicenter collaborative network and databases such as PubMed and EMBASE.

[0051] Perform quality control on raw data: Verify basic patient information, PDO culture conditions (culture medium, culture time), half-maximal inhibitory concentration determination method, and drug concentration units, and convert them to μM. Eliminate samples with culture failure, missing data, or significant methodological discrepancies to ensure data consistency.

[0052] Finally, a standardized database including patient ID, chemotherapy regimen, half-maximal inhibitory concentration value, and testing laboratory was established, and historical half-maximal inhibitory concentration values were extracted from the standardized database.

[0053] Establishing a standardized database can achieve data traceability (linking individual cases), group statistics (accurately calculating medians according to the protocol) and cross-center difference calibration (identifying laboratory operation bias) through structured fields. After quality control, it provides unified and reliable benchmark data for model construction, ensuring the representativeness and comparability of historical half-maximal inhibitory concentration values.

[0054] SPSS was used to calculate the half-maximal inhibitory concentration (HIPC) value of each corresponding chemotherapy regimen, namely the historical half-maximal inhibitory concentration (HIPC) value.

[0055] The final inhibitory concentration (IC50) of 5-FU was 2.2500 (95% CI, 1.4410 to 3.1820), the IC50 of 5-FU + OX was 1.2020 (95% CI, 1.007 to 1.8880), and the IC50 of 5-FU + IRI was 0.7578 (95% CI, 0.3915 to 1.1640).

[0056] Step S213: obtaining the median half inhibition concentration value of the drug chemotherapy according to the historical half inhibition concentration value.

[0057] In this example, historical 50% inhibitory concentration (IC50) values for each chemotherapy regimen (5-FU, 5-FU + OX, or 5-FU + IRI) were first extracted from a standardized database. Extreme values exceeding 1.5 times the interquartile range were removed and the data were sorted in ascending order. For each data set, if the sample size was odd, the middle value was taken as the median; if the sample size was even, the average of the two middle values was taken. The 95% confidence interval was simultaneously calculated using the percentile method. Ultimately, a median baseline value for each regimen was generated, providing a unified quantitative standard for standardizing individual patient 50% IC50 values.

[0058] Step S214, calculating the standardized half inhibition concentration value of the colorectal cancer patient according to the actually measured half inhibition concentration value and the median half inhibition concentration value.

[0059] The standardized half-maximal inhibitory concentration value satisfies the following formula: in, is the normalized half-maximal inhibitory concentration value, To measure the actual half-maximal inhibitory concentration value, is the median half inhibitory concentration value.

[0060] Wherein, the standardized relative tumor growth rate of the colorectal cancer patient obtained by using the PDOX model comprises: Step S221: Divide the immunodeficient mice corresponding to the PDOX model into an experimental group and a control group.

[0061] In this example, tumor diameters were first precisely measured using a vernier caliper to screen mice with uniform tumor volumes, thus avoiding experimental bias due to varying tumor growth stages. A sample size of 6-8 mice per group was determined based on a pre-experimental power analysis. The mice were randomly assigned to the experimental and control groups in a 1:1 ratio using a random number table to ensure that each group of mice was balanced in terms of baseline characteristics, including strain, age, and pathological stage of the patient from whom the PDOX model was derived. After grouping, an independent sample t-test was performed to verify that the mean tumor volumes of the two groups were not significantly different, ensuring the scientific nature of the grouping.

[0062] Step S222 measures the tumor volumes of the immunodeficient mice in the experimental group and the control group to obtain the initial tumor volume of the experimental group and the initial tumor volume of the control group.

[0063] In this embodiment, the tumor diameters of the immunodeficient mice are measured in step S221, the measurement data are saved, and the initial tumor volumes of the experimental group and the control group are calculated according to the ellipsoid volume formula.

[0064] Step S223 , administering the drug chemotherapy to the experimental group and administering a solvent or placebo of the same volume as the drug chemotherapy to the control group, to obtain the post-experiment tumor volume of the experimental group and the post-experiment tumor volume of the control group.

[0065] In this example, when drug chemotherapy was implemented in the experimental group and an equal volume of solvent or placebo was administered to the control group, a standardized dosing regimen was strictly followed: chemotherapy drugs were prepared according to the clinical equivalent dose (converted by the body surface area method, such as multiplying the recommended human dose by a conversion factor of 0.018 when converting it to the mouse dose), and the experimental group was administered by intraperitoneal injection, with a fixed cycle set (twice a week for 3 consecutive weeks); the control group was simultaneously injected with an equal volume of normal saline (water-soluble drugs) or a fat-soluble drug solvent containing less than or equal to 0.1% DMSO to ensure that the solvent component did not affect tumor growth.

[0066] During dosing, mice were housed in an SPF environment with free access to sterile feed and drinking water. Before each dosing, mice were weighed using an electronic balance, and the dosing volume was adjusted based on actual body weight to avoid dose bias. Tumor volume was measured twice weekly at fixed times after dosing. The longest and short perpendicular diameters of the tumor were measured using a vernier caliper with an accuracy of 0.01 mm, and the volume was calculated using the ellipsoid volume formula. All of these procedures should be performed by the same experimenter under the same lighting conditions to minimize measurement bias.

[0067] If any mouse in the experimental or control group developed a tumor volume of 2000 mm³ or exhibited signs of weakness, such as slowed movement or weight loss greater than 20%, the experiment was terminated and the final tumor volume recorded. A record sheet containing mouse number, group, dosing regimen, dosing time, dose, and post-experimental volume was established to ensure data traceability. This process ensured that the only variable between the experimental and control groups was whether or not they received drug intervention, through standardized dosing parameters, controlled housing conditions, and measurement procedures.

[0068] Step S224, calculating the standardized relative tumor growth rate of the colorectal cancer patient based on the initial tumor volume of the experimental group, the initial tumor volume of the control group, the post-experimental tumor volume of the experimental group, and the post-experimental tumor volume of the control group.

[0069] The normalized relative tumor proliferation rate satisfies the following formula: in, is the normalized relative tumor growth rate, is the tumor volume of tumor-bearing mice in the experimental group at the end of the experiment, is the tumor volume of tumor-bearing mice in the experimental group at the beginning of the experiment, is the tumor volume of tumor-bearing mice in the control group at the end of the experiment, Tumor volumes of tumor-bearing mice in the control group at the beginning of the experiment.

[0070] Step S3, obtaining the medical records and pathological examination of the colorectal cancer patient, and obtaining the patient's age, ASA score, Ki-67 index, combined positive score and clinical outcome based on the medical records and the pathological examination.

[0071] In this example, the patient's age can be directly extracted from the "Date of Birth" field on the first page of the medical record and calculated as of the date of surgery or diagnosis.

[0072] The ASA score was obtained by reviewing the preoperative evaluation records of the anesthesiology department and extracting the patient's preoperative physical condition score according to the American Society of Anesthesiologists (ASA) grading standard (grade I-V).

[0073] The Ki-67 index is obtained by querying the immunohistochemistry results section of the pathology diagnosis report and extracting the percentage of positive cells for the "Ki-67" item. If the report is not clearly marked, the pathology sections can be reviewed by two pathologists who independently interpret the results and take the average value.

[0074] The combined positive score was obtained by querying the pathology report of the PD-L1 test and extracting the CPS value in the "Immunohistochemistry-Tumor Microenvironment Analysis" section, which is the proportion of PD-L1-positive cells in tumor cells and tumor-infiltrating immune cells (lymphocytes and macrophages) to all tumor cells.

[0075] Data extraction follows a standardized process. For cases with missing fields in the electronic medical record, manual review of paper records is performed to supplement them. Pathological parameters are reviewed and confirmed by a certified pathologist to ensure that the interpretation of the Ki-67 index and CPS values conforms to the latest guidelines (such as ESMO or NCCN standards). The final result is a structured data table containing patient ID, age, ASA score, Ki-67 index, and combined positive score.

[0076] Clinical outcomes refer to the imaging examination reports of patients after chemotherapy included in the medical records. Doctors judge the effectiveness of chemotherapy (complete remission CR, partial remission PR, stable disease SD, progressive disease PD) by measuring changes in tumor size and evaluating tumor regression rate based on solid tumor efficacy evaluation criteria (such as RECIST standards). Combined with the changes in tumor cells in pathological examination (such as cell activity, morphological changes, etc.), the doctor finally defines the chemotherapy response result as "effective" (CR / PR) or "ineffective" (SD / PD), forming a binary clinical outcome.

[0077] Step S4, establishing a sample set using the half-maximal inhibitory concentration value, the standardized relative tumor proliferation rate, the patient's age, the ASA score, the Ki-67 index, the combined positive score, and the clinical outcome.

[0078] In this example, a sample set was constructed using multiple data sources, including 50% inhibitory concentration (IC), normalized relative tumor growth rate (NRTR), patient age, ASA score, Ki-67 index, and combined positive score, as independent variables, and clinical outcome as the dependent variable. Data correlation was required to ensure unique identification and correspondence between each data set. Continuous variables (IC, age, Ki-67 index, and combined positive score) were standardized: IC values were rounded to three significant figures (e.g., 2.250 μM), Ki-67 index and combined positive score were taken as the original percentage values from the pathology report (e.g., 30% and 15), and age was rounded to the nearest year. Categorical variables, such as ASA score, were numerically coded (grade I = 1, grade II = 2, grade III = 3, grade IV = 4, grade V = 5). Valid clinical outcomes were assigned a value of 1, and invalid clinical outcomes were assigned a value of 0.

[0079] Samples with missing key data such as unmeasured half-inhibitory concentration values or omissions in the CPS report, or logical contradictions (such as Ki-67 index greater than 100%) were eliminated to form a structured data set containing patient ID, half-inhibitory concentration, normalized relative tumor proliferation rate, age, ASA score, Ki-67 index, and combined positive score, where each sample row corresponds to one patient and the column field clearly defines the data type.

[0080] Finally, the structured dataset was normalized. Dimensionless continuous variables such as the 50% inhibitory concentration (PIC), age, Ki-67 index, and combined positive score were first screened. Extreme outliers such as PIC were identified using box plots and removed using the 95% quantile replacement method to minimize dimensionality interference. For variables with right-skewed distributions, such as PIC, a logarithmic transformation was performed to optimize data normality. Min-Max scaling was then used to reduce all continuous variables to the [0, 1] range. Dimensionless or standardized features, such as the standardized relative tumor growth rate (SRTR) and the American Society of Anesthesiologists (ASA) score, were retained as is to maintain physical meaning. Finally, the sample set was formed.

[0081] Step S5, constructing a model for predicting colorectal cancer drug chemotherapy response based on multivariate logistic regression, using the sample set to train the model for predicting colorectal cancer drug chemotherapy response, and evaluating the model for predicting colorectal cancer drug chemotherapy response after training.

[0082] In this embodiment, the sample set is first normalized. Specifically, normalizing the sample set means mapping feature data of different dimensions and distribution ranges (such as half-maximal inhibitory concentration values, patient age, Ki-67 index, etc.) to a unified numerical range through standardized conversion. The principle is to eliminate the disparity in the numerical range of the original features caused by dimensional differences, and avoid the model from being overly sensitive to features with larger values. The normalized sample set can improve the efficiency of model training, make the gradient descent process converge faster, and prevent feature weights from being affected by dimensions, ensuring that the model's judgment of the importance of all features is more objective. At the same time, it can also enhance the stability and generalization ability of the model and reduce the risk of overfitting due to data distribution differences.

[0083] Then, the sample set was divided into training set and test set using 7:3 stratified sampling. By combining forward / backward stepwise regression with grid search to optimize the regularization parameters, the best feature combination was selected and the model was fitted, and the regression coefficient was output. The final model for predicting the response of colorectal cancer chemotherapy drugs satisfies the following formula in, To predict the probability of chemotherapy resistance, is the normalized relative tumor growth rate, is the standardized half-maximal inhibitory concentration value, is the patient's age, For ASA rating, is the Ki-67 index, is the combined positive score.

[0084] The accuracy of the model was 91.11% (95% CI, 79.27% to 96.49%). Specifically, the sensitivity was 83.33% (95% CI, 55.20% to 97.04%), the specificity (1-false positive rate) was 93.94% (95% CI, 80.39% to 98.92%), and the positive and negative predictive values were 83.33% (95% CI, 55.20% to 97.04%) and 93.94% (95% CI, 80.39% to 98.92%), respectively.

[0085] The evaluation of the model for predicting colorectal cancer chemotherapy response after training specifically includes the following steps: Step S501 , extracting the clinical outcome of each sample in the test set, and using the model for predicting colorectal cancer drug chemotherapy response to obtain the predicted probability of each sample.

[0086] In this example, clinical outcome data for each sample was first extracted from the test set (effective = 1, ineffective = 0), representing the patient's actual chemotherapy response. Subsequently, the standardized half-maximal inhibitory concentration (SICC), relative tumor growth rate, and patient age from the test set were input into a trained model for predicting colorectal cancer chemotherapy response. The model calculated the log-odds ratio (LOR) for each sample based on the trained formula and converted it to a predicted probability between 0 and 1 using a sigmoid function. The predicted probability here refers to the probability of drug resistance, which indicates the likelihood that the sample will respond to chemotherapy.

[0087] Step S502 : setting multiple classification thresholds, comparing the predicted probability with the classification threshold, and calculating the sensitivity and false positive rate under the classification threshold based on the comparison result and the clinical outcome.

[0088] In this embodiment, first, a series of continuous classification thresholds are set in the probability range of 0 to 1 (for example, 20 thresholds such as 0.05, 0.10, 0.15...0.95 are selected with an interval of 0.05), and each threshold represents the critical value for converting the predicted probability into a valid or invalid decision.

[0089] For each classification threshold, the predicted probability of all samples in the test set is compared with the classification threshold: if the predicted probability is greater than or equal to the classification threshold, the prediction is judged to be valid and the predicted value is 1; otherwise, the prediction is judged to be invalid and the predicted value is 0.

[0090] Subsequently, a confusion matrix was constructed based on the clinical outcomes of the samples (effective as 1, ineffective as 0) and the prediction results of the model for predicting the response to colorectal cancer drug chemotherapy.

[0091] First, count four types of results: true positive is the number of samples that are actually effective and predicted to be effective, that is, the number of patients who are correctly identified to be effective in chemotherapy; false positives The number of samples that are actually ineffective but predicted to be effective, that is, patients who are mistakenly judged to be effective but ineffective for chemotherapy; false negative The number of samples that are actually effective but predicted to be ineffective, that is, the number of patients who are missed to receive chemotherapy; true negative is the number of samples that are actually ineffective and predicted to be ineffective, that is, the correctly identified chemotherapy-ineffective patients.

[0092] These four types of results constitute a 2×2 confusion matrix, with the horizontal representation of the model's prediction being valid or invalid, and the vertical representation of the actual clinical outcome being valid or invalid, clearly presenting the distribution of the model's correct predictions (TP, TN) and incorrect predictions (FP, FN).

[0093] According to the confusion matrix, calculate the sensitivity under the classification threshold , which reflects the ability of the model to predict the response to colorectal cancer chemotherapy to correctly identify effective patients. The calculation formula is: According to the confusion matrix, calculate the false positive rate under the classification threshold , which reflects the proportion of ineffective patients misjudged as effective by the model predicting the response to colorectal cancer chemotherapy. The calculation formula is: By traversing all classification thresholds, multiple sets of sensitivity and false positive rate data pairs are finally obtained.

[0094] Step S503 , with sensitivity as the vertical axis and false positive rate as the horizontal axis, the sensitivity and false positive rate under the classification threshold are used as coordinate points, and the coordinate points are plotted into a curve.

[0095] In this embodiment, the sensitivity and false positive rate calculated under different classification thresholds are used as coordinate points, which are sequentially plotted in a two-dimensional coordinate system and connected to form a curve.

[0096] Specifically, for each preset classification threshold, the corresponding sensitivity and false positive rate are calculated through the confusion matrix to form a set of coordinate data. Subsequently, the coordinate points corresponding to all classification thresholds are marked in the coordinate system in order from low to high classification thresholds. The horizontal axis is the false positive rate, ranging from 0 to 1, and the vertical axis is the sensitivity, ranging from 0 to. Finally, these coordinate points are connected by a smooth curve to form a complete curve, such as Figure 2 As shown, the curve intuitively demonstrates the classification performance of the model for predicting colorectal cancer drug chemotherapy response at different decision thresholds. The closer the curve is to the upper left corner (the higher the sensitivity and the lower the false positive rate), the stronger the ability of the model for predicting colorectal cancer drug chemotherapy response to distinguish between patients who are effective and those who are ineffective with chemotherapy.

[0097] Step S504: Calculate the area under the curve and its 95% confidence interval. In this embodiment, the trapezoidal method is used to calculate the area under the curve. The area enclosed by the curve and the coordinate axis is divided into multiple trapezoids. The area of each trapezoid is calculated in sequence, and the areas of all trapezoids are accumulated to obtain the area value, which quantifies the overall ability of the model to distinguish patients who respond to chemotherapy from those who do not.

[0098] The area value refers to the area under the curve, which is a core quantitative indicator used to measure the ability of the model for predicting colorectal cancer drug chemotherapy response to distinguish between effective and ineffective chemotherapy patients.

[0099] The Bootstrap method was used to calculate the 95% confidence interval of the area value. Sampling was repeated with replacement from the original sample set. The area value was recalculated each time to generate a distribution of area values. The value at the 2.5% position in the distribution was taken as the lower limit, and the value at the 97.5% position was taken as the upper limit, thus obtaining a 95% confidence interval, which reflects the fluctuation range of the estimated area value.

[0100] Step S505 : setting a reference threshold, comparing the difference between the area value and the reference threshold through a two-sided test, and calculating the significance level of the difference.

[0101] In this embodiment, first, a benchmark threshold is set, and the benchmark threshold is 0.5, which serves as a reference standard for judging whether the model for predicting colorectal cancer drug chemotherapy response has actual predictive value.

[0102] Subsequently, a two-sided statistical test was performed to compare the area under the curve (AUC) calculated by the model with a baseline threshold of 0.5. The null hypothesis for the two-sided test was that the AUC was equal to 0.5, indicating that the model's ability to predict colorectal cancer chemotherapy response was no different from random chance. The alternative hypothesis was that the AUC was not equal to 0.5, indicating that the model's ability to predict was significantly different from random chance.

[0103] By calculating the test statistic and combining it with the normal distribution or sampling distribution theory, we can derive a value reflecting the significance of the difference, namely the significance level. Specifically, we calculate the area value and its standard error based on the sample data. The standard error can be estimated using the bootstrap resampling method or the parametric method, reflecting the sampling error of the area value estimate. Subsequently, we construct the test statistic using the formula: in, The value is a measure of the deviation of the area value from the reference value relative to the multiple of the sampling error. is the area value, Baseline threshold, is the standard error.

[0104] Based on the central limit theorem, when the sample size is large, the sampling distribution of AUC approximately obeys the normal distribution, so the significance level can be derived using the standard normal distribution.

[0105] Based on a two-sided test, the significance level is defined as the absolute value of the standard normal distribution greater than The extreme value probability of is calculated as follows: is the probability of observing a difference between the current area value and 0.5 under the premise that the null hypothesis is true, is the cumulative distribution function of the standard normal distribution.

[0106] The significance level satisfies the following formula: in, is the significance level, is the cumulative distribution function of the standard normal distribution, is the area value, Baseline threshold, is the standard error.

[0107] Step S506: Comprehensively evaluate the model for predicting colorectal cancer drug chemotherapy response using the area value, the 95% confidence interval, and the significance level.

[0108] When the comprehensive evaluation satisfies the conditions that the area value is greater than 0.5, the lower limit of the 95% confidence interval is greater than 0.5, and the significance level is less than 0.05, the model for predicting colorectal cancer drug chemotherapy response is determined to be an available model.

[0109] In this embodiment, the area value reflects the ability of the model for predicting the response to chemotherapy of colorectal cancer drugs to distinguish between effective and ineffective patients. If the area value is greater than 0.5, it indicates that the predictive performance of the model for predicting the response to chemotherapy of colorectal cancer drugs is better than random guessing. The 95% confidence interval further measures the stability of the area value estimate. When the lower limit is greater than 0.5, it means that the true value of the area value has a 95% probability of being significantly higher than the random level, eliminating the possibility of the area value being inflated due to sampling error. The significance level verifies the model performance from a statistical point of view. If it is less than 0.05, it means that the effect of the model in distinguishing chemotherapy responses is not accidental and is statistically significant. Only when all three are met at the same time, that is, the area value is greater than 0.5, the lower limit of the 95% confidence interval is greater than 0.5, and the significance level is less than 0.05, can the model be determined to be an available model to ensure that the model for predicting the response to chemotherapy of colorectal cancer drugs has practical application value and reliability in predicting the response to chemotherapy of colorectal cancer drugs.

[0110] Step S6: Based on the evaluation results, the colorectal cancer drug chemotherapy response model is used to provide chemotherapy guidance for the colorectal cancer patient to be treated.

[0111] The method of using the colorectal cancer drug chemotherapy response model to provide chemotherapy guidance for colorectal cancer patients specifically includes the following sub-steps: Step S601: obtaining individualized data of a rectal cancer patient to be treated, and inputting the individualized data into the model for predicting the response to rectal cancer chemotherapy to obtain the probability of drug resistance to the chemotherapy.

[0112] In this embodiment, the individualized data includes patient age, ASA score, Ki-67 index, combined positive score, normalized relative tumor growth rate, and normalized half-maximal inhibitory concentration. The method for obtaining the individualized data is exactly the same as the above method and will not be repeated here.

[0113] Step S602 : setting a drug resistance probability threshold, comparing the drug resistance probability with the drug resistance probability threshold, and determining, based on the comparison result, that the rectal cancer patient to be treated is a chemotherapy-sensitive population.

[0114] In this embodiment, the drug resistance probability threshold is verified based on a large amount of data from a model for predicting colorectal cancer drug chemotherapy response and is obtained in combination with clinical practice. The drug resistance probability threshold is 0.5, which serves as a key dividing point for distinguishing between chemotherapy sensitivity and drug resistance in patients.

[0115] Next, the model outputs the drug resistance probability of the rectal cancer patient to be treated, and then accurately compares it with the set resistance probability threshold. If the patient's resistance probability is less than the resistance probability threshold, the patient is clearly determined to be chemotherapy-sensitive, indicating a high probability of an effective response to conventional chemotherapy regimens, and the standard chemotherapy strategy can be prioritized in clinical planning. Conversely, if the resistance probability is greater than or equal to the threshold, the patient is classified as a drug-resistant risk group, requiring further exploration of individualized alternative treatments or combination therapies. This quantitative comparison process provides clear and actionable criteria for the clinical precision screening of chemotherapy-sensitive rectal cancer patients, effectively supporting the formulation of individualized treatment decisions and improving the targetedness and effectiveness of treatment plans.

[0116] Step S603: Based on the fact that the rectal cancer patient to be treated is a chemotherapy-sensitive population, the individualized data is combined to determine the physical tolerance of the rectal cancer patient to be treated to the drug chemotherapy.

[0117] In this example, the patient's physiological indicators are first extracted, including age, physical status score, and comorbidity index. Next, laboratory test results are analyzed, focusing on bone marrow reserve function, liver and kidney function, and nutritional status. Next, genetic and molecular features are integrated. Finally, combined with tumor biological behavior, a tolerance assessment model is constructed using multidimensional data. A weighted scoring system or machine learning algorithm is used to quantify the patient's tolerance to chemotherapy drug dose intensity and number of cycles, that is, the patient's physical tolerance to chemotherapy drugs.

[0118] Step S604: determining the cycle and dosage of the drug chemotherapy according to the body's tolerance.

[0119] Based on the quantified results of the physical tolerance assessment, combined with the tumor treatment guidelines and drug instructions, the cycle and dose of drug chemotherapy are determined in steps: First, the preset dose adjustment rules are applied according to the tolerance level. Next, the number of cycles is determined based on the tumor stage and tolerance. For example, the standard cycle for stage III patients is 6-8 cycles. Those with high tolerance use a full 8 cycles, those with medium tolerance are adjusted to 6 cycles, and those with low tolerance are further shortened to 4 cycles and the frequency of efficacy evaluation is increased. Finally, a dynamic correction mechanism is introduced to monitor the patient's actual toxic reactions after the first cycle. Ensure that the chemotherapy regimen is in line with the patient's individual tolerance and follows the principle of sufficient dose and sufficient cycle of anti-tumor treatment, so as to achieve a precise balance between maximizing efficacy and safety.

[0120] like Figure 3 As shown, on the other hand, the present invention also provides a colorectal cancer drug chemotherapy response prediction system, including an input device, a processor, an output device and a memory, wherein the input device, processor, output device and memory are interconnected, the memory includes the computer-readable storage medium mentioned above, the memory is used to store a computer program, the computer program includes program instructions, and the processor is configured to call the program instructions.

[0121] In this embodiment, the input device is used to provide input-related data or instructions to the system. In the colorectal cancer chemotherapy response prediction system, the input device can include common human-computer interface devices such as a keyboard, mouse, and touch screen. Through the input device, doctors or researchers can enter necessary parameters.

[0122] The processor is the core component of the system, responsible for executing computer program instructions and performing data processing and analysis. In the colorectal cancer chemotherapy response prediction system, the processor analyzes and interprets input experimental data by running pre-programmed algorithms and models. The processor can be a central processing unit (CPU), a graphics processing unit (GPU), or other dedicated processing unit.

[0123] The memory is used to store computer programs, data, and parameters required by the system. It can include random access memory (RAM) for temporary data storage and processing, and persistent memory (such as a hard disk or solid-state drive) for long-term storage and preservation of data.

[0124] The output device is used to present the results of system processing and analysis to a user or external device. The output device can be a display, printer, chart drawing device, etc. Through the output device, the system can display the prediction results for reference by doctors, researchers, or patients to assist in decision-making and communication.

[0125] Compared with single PDX model and PDO model: For colorectal cancer samples, the PDOX model has a significantly higher success rate than traditional PDX models, and its modeling is much simpler. The PDOX model is primarily constructed by transplanting in vitro cultured organoids into immunodeficient mice via injection. This avoids the often-invasive anesthesia and surgical procedures commonly used in PDX models, reducing the risk of surgical mortality in mice. Furthermore, the PDOX model also takes longer to develop tumors than the PDX model, further improving research efficiency.

[0126] The PDO model primarily simulates tumor cell growth and drug response in vitro and has high clinical relevance. However, its in vivo validation is relatively insufficient, limiting its application in clinical translation. The PDOX model component of the combined prediction model provides a basis for in vivo validation to address this issue. By testing drugs in immunodeficient mice, the PDOX model can more accurately reflect drug efficacy and patient response in the in vivo environment, thus addressing the inadequacy of in vivo validation of the PDO model alone.

[0127] A single PDO or PDX model can be unstable in certain situations, for example, results may vary between laboratories or under different operating conditions. By integrating multiple variables and models, a combined prediction model can reduce the errors and uncertainties associated with a single model, improving the stability and reliability of predictions. This multi-model approach not only leverages the strengths of each model but also overcomes the limitations of a single model, providing a more comprehensive and accurate platform for cancer research and drug screening.

[0128] Compared with the PDO-PDX combined model: Comprehensive Inclusion of Individual Patient Characteristics: The PDO-PDX combined model primarily focuses on simulating tumor drug responses both in vitro and in vivo, but relatively little consideration is given to individual patient characteristics. The final combined prediction model incorporates not only relevant variables from the PDO and PDX models, but also variables related to the patient's general condition and health status, such as age and ASA score. This allows for a more comprehensive consideration of the impact of individual patient differences on chemotherapy response, thereby improving the accuracy and practicality of predictions.

[0129] Richer prediction dimensions: The combined prediction model not only predicts the direct effects of drugs on tumor cells but also predicts chemotherapy response from multiple dimensions, taking into account the patient's overall health status (such as ASA score), tumor cell proliferation activity (such as Ki-67 index), and immune-related indicators (such as PD-L1 expression level). This makes the prediction results richer and more comprehensive, providing a more comprehensive reference for clinical treatment.

[0130] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some or all of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and description of the present invention.

Claims

1. A computer-readable storage medium, characterized in that The computer readable storage medium stores a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by a processor, the processor is caused to perform the following steps: Obtaining tumor cells from colorectal cancer patients, and constructing a PDO model and a PDOX model based on the tumor cells; Based on drug chemotherapy, the PDO model is used to obtain the standardized half-maximal inhibitory concentration value of the colorectal cancer patient, and the PDOX model is used to obtain the standardized relative tumor growth rate of the colorectal cancer patient; Obtaining the medical records and pathological examinations of the colorectal cancer patients, and obtaining the patient's age, ASA score, Ki-67 index, combined positive score, and clinical outcome based on the medical records and pathological examinations; establishing a sample set using the half-maximal inhibitory concentration value, the standardized relative tumor proliferation rate, the patient's age, the ASA score, the Ki-67 index, the combined positive score, and the clinical outcome; constructing a model for predicting colorectal cancer drug chemotherapy response based on multivariate logistic regression, training the model for predicting colorectal cancer drug chemotherapy response using the sample set, and evaluating the model for predicting colorectal cancer drug chemotherapy response after training; Based on the results of the evaluation, the colorectal cancer drug chemotherapy response model is used to provide chemotherapy guidance for patients with colorectal cancer.

2. A computer-readable storage medium according to claim 1, characterized in that The construction of the PDO model and the PDOX model based on the tumor cells comprises: The tumor cells are cultured in three dimensions in vitro to construct a PDO model that retains the pathological, genetic characteristics, and drug response properties of the tumor cells; The PDO model was transplanted into immunodeficient mice to construct a PDOX model that simulates the in vivo tumor microenvironment.

3. The computer-readable storage medium according to claim 1, wherein: The method of obtaining the standardized half-maximal inhibitory concentration value of the colorectal cancer patient using the PDO model includes: Conducting chemotherapy experiments with different concentrations of the drug on the PDO model to obtain the actual measured half-maximal inhibitory concentration value of the colorectal cancer patient; Collect historical half-inhibitory concentration values for colorectal cancer patients treated with the same chemotherapy drug; Obtaining a median half inhibition concentration value of the chemotherapy drug according to the historical half inhibition concentration value; The standardized half inhibition concentration value of the colorectal cancer patient is calculated according to the actual measured half inhibition concentration value and the median half inhibition concentration value.

4. A computer-readable storage medium according to claim 3, characterized in that: The standardized half-maximal inhibitory concentration value satisfies the following formula: in, is the normalized half-maximal inhibitory concentration value, To measure the actual half-maximal inhibitory concentration value, is the median half inhibitory concentration value.

5. The computer-readable storage medium according to claim 1, wherein: The method of obtaining the standardized relative tumor growth rate of the colorectal cancer patient using the PDOX model includes: The immunodeficient mice corresponding to the PDOX model were divided into an experimental group and a control group; Measuring the tumor volumes of the immunodeficient mice in the experimental group and the control group to obtain the initial tumor volume of the experimental group and the initial tumor volume of the control group; The experimental group is administered the chemotherapy drug, and the control group is administered a solvent or placebo of the same volume as the chemotherapy drug, to obtain the post-experiment tumor volume of the experimental group and the post-experiment tumor volume of the control group; The standardized relative tumor growth rate of the colorectal cancer patients is calculated according to the initial tumor volume of the experimental group, the initial tumor volume of the control group, the post-experimental tumor volume of the experimental group and the post-experimental tumor volume of the control group.

6. The computer-readable storage medium according to claim 1, wherein: The sample set includes a test set, and the evaluation of the model for predicting colorectal cancer drug chemotherapy response after training includes: Extracting the clinical outcome of each sample in the test set, and using the model for predicting colorectal cancer drug chemotherapy response to obtain a predicted probability for each sample; Setting multiple classification thresholds, comparing the predicted probability with the classification threshold, and calculating the sensitivity and false positive rate under the classification threshold based on the comparison result and the clinical outcome; With sensitivity as the vertical axis and false positive rate as the horizontal axis, the sensitivity and false positive rate under the classification threshold are used as coordinate points, and the coordinate points are plotted into a curve; Calculate the area under the curve and the 95% confidence interval of the area; Setting a benchmark threshold, comparing the difference between the area value and the benchmark threshold through a two-sided test, and calculating the significance level of the difference; The area value, the 95% confidence interval and the significance level are used to comprehensively evaluate the model for predicting colorectal cancer drug chemotherapy response.

7. The computer-readable storage medium according to claim 6, wherein: When the comprehensive evaluation satisfies the conditions that the area value is greater than 0.5, the lower limit of the 95% confidence interval is greater than 0.5, and the significance level is less than 0.05, the model for predicting colorectal cancer drug chemotherapy response is determined to be an available model.

8. The computer-readable storage medium according to claim 1, wherein: The model for predicting colorectal cancer chemotherapy response satisfies the following formula: in, To predict the probability of chemotherapy resistance, is the normalized relative tumor growth rate, is the standardized half-maximal inhibitory concentration value, is the patient's age, For ASA rating, is the Ki-67 index, is the combined positive score.

9. The computer-readable storage medium according to claim 1, wherein: The method of using the colorectal cancer drug chemotherapy response model to provide chemotherapy guidance for colorectal cancer patients includes: Obtaining individualized data of a rectal cancer patient to be treated, and inputting the individualized data into the model for predicting rectal cancer drug chemotherapy response to obtain a probability of drug resistance to the drug chemotherapy; Setting a drug resistance probability threshold, comparing the drug resistance probability with the drug resistance probability threshold, and determining, based on a result of the comparison, that the rectal cancer patient to be treated is a chemotherapy-sensitive population; Based on the fact that the rectal cancer patient to be treated is a chemotherapy-sensitive population, the physical tolerance of the rectal cancer patient to be treated to the drug chemotherapy is determined in combination with the individualized data; The cycle and dosage of the drug chemotherapy are determined according to the body's tolerance.

10. A colorectal cancer chemotherapy response prediction system, characterized in that: The invention comprises an input device, a processor, an output device and a memory, wherein the input device, the processor, the output device and the memory are connected to each other, the memory comprises a computer-readable storage medium according to any one of claims 1 to 9, the memory is used to store a computer program, the computer program comprises program instructions, and the processor is configured to call the program instructions.

Images